## 1. Model Based Simulations and the Actual Data

## Source details

**Canonical URL:** [1. Model Based Simulations and the Actual Data](https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025073.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025073.pdf.md)
- [Structured JSON version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025073.pdf.json)

---

### A. Purpose and overview
- Objective: deepen Fund engagement in macrofinancial surveillance in the WAEMU by:
  - analyze unobserved credit cycle dynamics and risks;
  - emphasize macrofinancial linkages and macroprudential policy;
  - examine interactions between monetary and macroprudential policies;
  - generate medium-term projections and alternative scenarios; and
  - assess feedbacks between credit, the real economy, and bank solvency.
- Prepared by Knarik Ayvazyan (MCMSR) with suggestions from Luca Antonio Ricci and Lawrence Norton.
- A detailed description will be presented in a forthcoming IMF working paper.
- Publication date: April 17, 2025.

### B. Model description (structure and assumptions)
- Modeling approach:
  - Semi-structural model with New Keynesian features (monopolistic competition, nominal and real rigidities, non-neutral short-run monetary policy).
  - Behavioral equations interpreted economically but estimated as reduced-form coefficients rather than deep DSGE parameters.
  - All real variables decomposed into trend and gap components (deviation from long-run trend).
- Blocks of the model:
  - Domestic: aggregate demand, aggregate supply (Phillips curve), uncovered interest parity (UIP), credit cycle, bank capital buffers, monetary policy rule.
  - Foreign economy: exogenous AR(1) processes using advanced economies data; steady-state values reflect sample averages.
- Expectations: formed based on lagged effects and model-predicted future outcomes.

### C. Credit cycle formulation and bank capital buffers (innovations)
- Credit gap dynamics (Equations 1–2):
  - Equation (1): CC_gap,t = i1 CC_gap,t−1 + i2 Y_gap,t + epsilon_c,t
    - 퐶퐶_푔푎푝,푡 is the real credit gap (private non-financial sector credit minus its trend).
    - 푌_푔푎푝 is the output gap.
    - 휖_c,t is a credit-supply shock unrelated to aggregate demand and past credit adjustments.
  - Equation (2): 휖_c,t = i3 (BBBt − BBB_ss) + ε_c,t
    - BBB_t is capital buffer (capital ratio relative to supervisory threshold).
    - BBB_ss is steady-state level of capital buffers (calibrated as historical long-term average).
    - ε_c,t is white-noise shocks to real credit.
- Feedback from credit to real economy:
  - Credit channel operates through bank capital constraints and costly equity issuance (failure of Modigliani-Miller assumptions).
  - Undercapitalized banks may cut lending to restore capital ratios, potentially reducing economic growth.
- Bank capital buffers dynamics (Equations 3–5):
  - B_t = K_t − K_t^*
  - B_t = ρ_B B_t−1 + (1−ρ_B) B_ss + i4 (C_t^g − C_ss^g) + ε_B,t
  - Regulatory requirement K_t^* follows a random walk: K_t^* = K_t−1^* + ε_K^*
- Role of macroprudential policy:
  - Regulatory changes to capital requirements are the primary macroprudential tool in WAEMU but have been little used.
  - The countercyclical capital buffer (CCyB) exists but "has been set to zero since its creation".
  - Tightening macroprudential stance or higher capital requirements can lower capital buffers, reduce credit supply, and dampen credit and real cycles.
- International and empirical context:
  - References to empirical findings (e.g., Jimenez et al. (2009), Aiyar et al. (2014), Adrian and Shin (2010), BCBS Principle 16) supporting capital–lending linkages and benefits of releasable buffers.
  - Empirical WAEMU institutional note: CSF-UMOA and BCEAO’s Macroprudential Policy Committee have made limited reliance on macroprudential measures; CSF-UMOA has rarely issued recommendations on implementation.

### D. Aggregate demand and credit interaction
- Aggregate demand equation (Equation 6) links output gap to:
  - its lag (β1 Y_gap,t−1),
  - monetary condition index (MC_M, with weight −β2),
  - foreign output gap (β3 Y*_gap,t),
  - credit shocks (β5 ε_c,t),
  - aggregate demand shock (ε_Y,t).
- Monetary condition index (Equation 7):
  - MC_M_t = β4 (R_real_gap_t + RISK_premium_t) + (1−β4)(−Z_gap_t)
  - RISK premium is negatively correlated with deviation of currency coverage ratio from trend.
- Interpretation:
  - Credit shocks ε_c,t allow modeling of feedback from credit-supply events (credit crunch) into aggregate demand.
  - An expansion of private credit unrelated to demand and past credit adjustments increases aggregate demand.

### E. Conventional model components: Phillips curve, UIP, and policy reaction function
- Phillips curve (Equations 8–9):
  - ∆CPI_t = α1 ∆CPI_t−1 + (1−α1) ∆CPI_t+1 + α2 RMC_t + ε_CPI,t
  - RMC_t = α3 Y_gap,t + (1−α3) Z_gap,t
- Uncovered interest parity (UIP) and exchange rate (Equation 10):
  - General UIP allows for degrees of exchange rate flexibility and capital controls.
  - S_t dynamics include lagged S, expected S_t+1, domestic and foreign nominal interest rates, and a UIP shock ε_S,t.
  - For WAEMU: coefficients assume exchange rate rigidity (h2 close to unity), capital controls (P1), desired long-term exchange variation and long-term real exchange variation set to zero (∆S^���� = 0 and ∆Z^���� = 0), consistent with PPP and the historical peg of the CFAF.
- Monetary policy rule and reaction function (Equations 11–13):
  - Short-term nominal interbank rate RS_t responds to:
    - 4-quarter-ahead inflation forecast deviations from target,
    - output gap,
    - deviation of risk premium from steady-state (influenced by reserve coverage).
  - Parameter h1 captures degree of interest rate control under the peg (h1 = 1: CB loses control; h1 = 0: full sterilization and control).
  - RS_t depends on its lag, neutral nominal interest rate RS_n, inflation expectations ∆4CPI_t+4, and a policy shock ε_RS,t.
  - Risk premium evolution:
    - PPP_t = d1 PPP_t−1 + (1−d1) PPP_ss − d2 CCR_gap_t + ε_PPP,t
    - CCR gap (currency coverage ratio deviation) evolves with k1 CCR_gap_t−1 + k2 (RS_t − RS_n) + k3 |Z_gap_t| + ε_CCR,t
  - The CCR long-term value chosen at 68 percent (chosen close to value at end 2024).

### F. Model parametrization and calibration
- Parameters and steady-state calibrations:
  - Capital buffer steady-state calibrated as historical long-term average.
  - CCR long-term value set at 68 percent.
  - Regulatory requirement K_t^* modeled as a random walk without drift.
- Empirical calibration draws on prior work (Carlos et al., 2022) and related literature; detailed parameter table is provided in Annex I (not reproduced here).

### G. Key findings and policy-relevant implications (from the model text)
- Model capabilities:
  - Decomposes real variables into trend and gap, captures unobserved credit cycle dynamics, and produces medium-term projections and scenarios.
  - Enables assessment of consistency between real and credit cycles and of macrofinancial trade-offs.
- Policy implications:
  - Banks transmit macroprudential policy to the real economy via capital buffers → credit supply → output.
  - Maintaining positive releasable capital buffers (e.g., CCyB) provides headroom and can mitigate procyclical lending cuts during downturns.
  - WAEMU’s CCyB has been set to zero since creation; increasing and maintaining releasable buffers could improve resilience.
  - Given the peg and reserve considerations, BCEAO’s monetary policy reacts to reserve/CCR developments; sterilization and capital controls affect effective policy autonomy.
- Institutional notes:
  - Prudential framework (established in 2018) allows BCEAO to implement macroprudential measures in line with financial stability mandate (Article 9 of BCEAO Statutes).
  - Since 2010 CSF-UMOA has rarely issued recommendations on macroprudential instrument implementation; committees have not always met prescribed frequencies.

### H. Parameterization and estimation approach (Bayesian mix)
- Estimation strategy:
  - Business cycle part relies on calibration similar to Carlos et al. (2022).
  - Newly introduced financial-block parameters: calibrated using regression analysis, then estimated via Bayesian techniques.
  - Structural parameters estimated with Bayesian estimation using observable variables over 2004Q1 to 2024Q3.
  - Bayesian steps: Kalman filtering to construct likelihood; combine priors with data; maximize log posterior to estimate mode; Metropolis-Hastings to obtain full posterior distributions and evaluate marginal likelihood.
- Calibration of steady-state and shock variances:
  - Steady-state parameters reflect target levels or historical averages.
  - Standard deviations of shocks generally calibrated unless explicitly estimated.
  - Calibration aims for reasonable dynamics of gaps and trends when filtering historical data.
- Out-of-sample forecasting:
  - Forecast exercise begins in 2010Q1.
  - Assumption: all foreign and unobserved variables (based on full-sample information) are observed throughout the forecast period.
  - All other variables are known until the quarter preceding each forecast realization over the next 8 quarters.
  - Figure 1 indicates good forecasting capabilities when shocks are not large.

### I. Historical decompositions, impulse responses, and projections
- Business cycle decomposition highlights:
  - Output gap negative during 2011 (Ivorian crisis and Sahel drought), moved around zero, rose until the COVID-2019 pandemic, and turned negative during the pandemic.
  - Post-pandemic recovery driven by favorable monetary conditions, gradual recovery of foreign output, and credit demand.
  - Large fiscal package in 2020 shows as large positive shocks in mid-2020.
  - Feedback from financial sector shocks to the business cycle is limited historically.
- Credit gap decomposition:
  - Positive credit gap narrowed since late 2023 and projected to become somewhat negative by end-2024.
  - Positive contributions from favorable business cycle outweighed by strong inertia of the credit gap, capital buffers below steady-state, and credit shocks.
  - Capital buffers above steady-state and persistence of credit were major contributors to positive credit gap from 2010 to 2018.
- Impulse response key results (Figure 4):
  - Positive demand shock:
    - Raises output gap and rate of inflation.
    - Boosts real credit growth by 1.6%.
    - Contributes to capital accumulation via increased profitability.
    - Calls for increase in real interest rate (nominal rate hike greater than rise in inflation); tightening dampens output and reduces credit demand; medium-term return to potential.
  - Positive credit shock:
    - Allows banks to accumulate capital via higher earnings from above-trend credit growth.
    - Tightening monetary conditions and regulatory discretionary actions put downward pressure on credit supply; credit growth turns negative after one quarter.
    - Credit shocks do not directly affect inflation in the model.
    - Higher capital requirement lowers additional capital buffer difference between bank-held capital and regulatory threshold, raises marginal lending costs, and reduces credit volume.
    - Impact on output is mild; output gap turns slightly negative, reflecting weak credit–real economy link due to limited financial deepening.
    - Results imply a relatively easy trade-off for macroprudential regulator when responding to excessive credit growth and systemic risk.
- Model-based projections and conditioning assumptions:
  - Projections conditioned on foreign inflation, output growth, and interest rates (IMF WEO October 2024 projections referenced for conditioning).
  - Model omits full external sector and expected developments in hydrocarbon production, current account improvements, and reserve accumulation; a CCR target close to end-2024 values was chosen to avoid a monetary policy reaction to CCR.
  - Projection highlights:
    - Macroeconomic conditions would call for normalization of monetary policy as FX reserve position improved in late 2024, inflation expectations decline, and credit gap is negative.
    - Negative credit gap expected to close gradually through forecasting period.
    - Credit demand expected to recover in line with activity with a lag; credit-to-GDP ratio to grow again as financial deepening resumes.
    - Bank profitability to improve and gradually restore capital buffers, but capital buffers remain below historical long-term average.
    - Inflation (quarterly) projected to converge to the target level by in early 2025.

### J. Policy implications and recommendations
- Model links macrofinancial shocks, cyclical risk, and capital buffers — useful for capital requirement policy discussions.
- Current historically low capital buffers relative to steady-state may be insufficient to withstand adverse domestic and international shocks.
- Recent financial soundness indicators and vulnerabilities:
  - Overall capital adequacy ratio declined from 13.7 percent at end-2023 to 13 percent in June 2024 (regulatory norm 11.5 percent).
  - NPLs and restructured loans increased from 8.7 percent at end-2023 to 9 percent in June 2024.
  - Liquidity conditions deteriorated.
- Recommendation: further strengthen capital buffers to mitigate risk amplifiers (concentration, contagion) and emerging risks (interest rate risk), given sovereign exposures and reliance on BCEAO refinancing.
- WAEMU policy actions:
  - WAEMU Council of Ministers decision (December 2023) to double minimum paid-in capital for banks to CFAF 20 billion will help system resilience.
- Financial deepening and central bank financing risks:
  - Banks near regulatory minimum face choice: restrict credit or increase capital buffers; restriction could hinder growth and investment.
  - Current credit penetration is relatively low, at about one-fourth of GDP.
  - Baseline model projects gradual recovery of capital buffers toward steady-state, suggesting financial deepening could be limited under current capital projections.
  - Heavy reliance on central bank refinancing poses risks; greater BCEAO financing to achieve deepening could increase financial stability risks for BCEAO.
- Macroprudential recommendations:
  - BCEAO should be ready to adjust macroprudential policy in line with financial cycle developments.
  - Consider moving toward a positive neutral level for the CCyB and develop a CCyB decision framework in parallel.
  - Gradual phase-in of higher capital requirements during financial expansion can mitigate excessive credit growth and strengthen banks’ loss-absorption capacity.
  - Use the model for simulation exercises to assess macrofinancial impacts of micro- and macroprudential capital requirement changes.

### K. Model usefulness, extensibility, and applications
- Model formalizes linkages between real and financial sectors, provides realistic forecasts, and helps assess risks and support monetary and financial policy analysis.
- Uses:
  - Analyze historical macroeconomic and financial sector data.
  - Interpret current disequilibria.
  - Conduct scenario analyses and examine empirical relevance of frictions.
  - Make forecasts to inform policy decisions.
- Extensibility:
  - Financial block intentionally simple; model is flexible for future extensions.
  - With minor modifications and additional data, model can address sector-specific financial risks.
  - Can be tailored to resemble an FSAP stress test using supervisory data.
  - Can evaluate macroprudential measures such as the CCyB.

### L. Annex I — Selected estimated and calibrated parameter values (as presented)
- Estimated Parameters (Prior distribution; Posterior Parameters Value):
  - 훼1: 훽(0.5,0.04) → 0.45
  - 훼2: N(0.45,0.04) → 0.43
  - 훼3: 훽(0.25,0.02) → 0.24
  - 훽1: 훽(0.72,0.06) → 0.69
  - 훽2: N(0.15,0.01) → 0.14
  - 훽3: N(0.28,0.03) → 0.26
  - 훽4: 훽(0.4,0.04) → 0.39
  - 훽5: N(0.1,0.01) → 0.1
  - 훾1: 훽(0.4,0.04) → 0.51
  - ℎ1: 훽(0.4,0.04) → 0.32
  - i1: 훽(0.65,0.06) → 0.69
  - i2: N(0.4,0.03) → 0.41
  - i3: N(0.3,0.03) → 0.3
  - i4: N(0.02,0.001) → 0.02
  - k1: 훽(0.94,0.1) → 0.98
  - k2: N(0.6,0.04) → 0.62
  - k3: N(0.35,0.04) → 0.34
  - 훽(0.9,0.09) → 0.94 (label in source: 휌휌_BBBBBB symbol)
- Calibrated Parameters (as listed in Annex):
  - h2 → 0.97
  - P1 → 0.5
  - g2 → 1
  - g3 → 1
  - g4 → 1
  - ∆S_t^����� → 0
  - ∆Z_t^����� → 0
  - ∆4CCPM_t^��������� → 2
  - ∆4CCPM_t^*����������� → 2
  - PPPPP_sess → 2
  - CC_ssg → 7
  - BBBBBB_sess → 2
  - RSSS_sess1 → 5
  - CCR_sess → 68
  - GGGPP_gPGggh_sess → 5.7

*Source — IMF Selected Issues Paper chapter "1. Model Based Simulations and the Actual Data" (sipea2025073).*

### 1. Model Based Simulations and the Actual Data ________________________________________ 8

### 1. Model Based Simulations and the Actual Data

### A. Purpose and overview
- This quantitative model for Macrofinancial Policy Analysis for the WAEMU was developed to deepen the Fund’s engagement in macrofinancial surveillance in the region and to:
  - analyze unobserved credit cycle dynamics and risks;
  - emphasize macrofinancial linkages and macroprudential policy;
  - examine interactions between monetary and macroprudential policies;
  - generate medium-term projections and alternative scenarios; and
  - assess feedbacks between credit, the real economy, and bank solvency.
- Prepared by Knarik Ayvazyan (MCMSR) with suggestions from Luca Antonio Ricci and Lawrence Norton. A detailed description will be presented in a forthcoming IMF working paper.
- Publication date: April 17, 2025.

### B. Model description (structure and assumptions)
- Modeling approach:
  - Semi-structural model with New Keynesian features (monopolistic competition, nominal and real rigidities, non-neutral short-run monetary policy).
  - Behavioral equations interpreted economically but estimated as reduced-form coefficients rather than deep DSGE parameters.
  - All real variables decomposed into trend and gap components (deviation from long-run trend).
- Blocks of the model:
  - Domestic: aggregate demand, aggregate supply (Phillips curve), uncovered interest parity (UIP), credit cycle, bank capital buffers, monetary policy rule.
  - Foreign economy: exogenous AR(1) processes using advanced economies data; steady-state values reflect sample averages.
- Expectations: formed based on lagged effects and model-predicted future outcomes.

### C. Credit cycle formulation and bank capital buffers (innovations)
- Credit gap dynamics (Equations 1–2):
  - Equation (1): CC_gap,t = i1 CC_gap,t−1 + i2 Y_gap,t + epsilon_c,t
    - 퐶퐶_푔푎푝,푡 is the real credit gap (private non-financial sector credit minus its trend).
    - 푌_푔푎푝 is the output gap.
    - 휖_c,t is a credit-supply shock unrelated to aggregate demand and past credit adjustments.
  - Equation (2): 휖_c,t = i3 (BBBt − BBB_ss) + ε_c,t
    - BBB_t is capital buffer (capital ratio relative to supervisory threshold).
    - BBB_ss is steady-state level of capital buffers (calibrated as historical long-term average).
    - ε_c,t is white-noise shocks to real credit.
- Feedback from credit to real economy:
  - Credit channel operates through bank capital constraints and costly equity issuance (failure of Modigliani-Miller assumptions).
  - Undercapitalized banks may cut lending to restore capital ratios, potentially reducing economic growth.
- Bank capital buffers dynamics (Equations 3–5):
  - B_t = K_t − K_t^*
  - B_t = ρ_B B_t−1 + (1−ρ_B) B_ss + i4 (C_t^g − C_ss^g) + ε_B,t
  - Regulatory requirement K_t^* follows a random walk: K_t^* = K_t−1^* + ε_K^*
- Role of macroprudential policy:
  - Regulatory changes to capital requirements are the primary macroprudential tool in WAEMU but have been little used.
  - The countercyclical capital buffer (CCyB) exists but "has been set to zero since its creation".
  - The model highlights that tightening macroprudential stance or higher capital requirements can lower capital buffers, reduce credit supply, and dampen credit and real cycles.
- International and empirical context:
  - References to empirical findings (e.g., Jimenez et al. (2009), Aiyar et al. (2014), Adrian and Shin (2010), BCBS Principle 16) supporting capital–lending linkages and benefits of releasable buffers.
  - Empirical WAEMU institutional note: CSF-UMOA and BCEAO’s Macroprudential Policy Committee have made limited reliance on macroprudential measures; CSF-UMOA has rarely issued recommendations on implementation.

### D. Aggregate demand and credit interaction
- Aggregate demand equation (Equation 6) links output gap to:
  - its lag (β1 Y_gap,t−1),
  - monetary condition index (MC_M, with weight −β2),
  - foreign output gap (β3 Y*_gap,t),
  - credit shocks (β5 ε_c,t),
  - aggregate demand shock (ε_Y,t).
- Monetary condition index (Equation 7) equals:
  - MC_M_t = β4 (R_real_gap_t + RISK_premium_t) + (1−β4)(−Z_gap_t)
  - RISK premium is negatively correlated with deviation of currency coverage ratio from trend.
- Interpretation:
  - Credit shocks ε_c,t allow modeling of feedback from credit-supply events (credit crunch) into aggregate demand.
  - An expansion of private credit unrelated to demand and past credit adjustments increases aggregate demand.

### E. Conventional model components: Phillips curve, UIP, and policy reaction function
- Phillips curve (Equations 8–9):
  - ∆CPI_t = α1 ∆CPI_t−1 + (1−α1) ∆CPI_t+1 + α2 RMC_t + ε_CPI,t
  - RMC_t = α3 Y_gap,t + (1−α3) Z_gap,t
  - Coefficients capture inflation persistence, pass-through from real marginal costs, and slope of Phillips curve.
- Uncovered interest parity (UIP) and exchange rate (Equation 10):
  - General UIP allows for degrees of exchange rate flexibility and capital controls.
  - S_t dynamics include lagged S, expected S_t+1, domestic and foreign nominal interest rates, and a UIP shock ε_S,t.
  - For WAEMU: coefficients assume exchange rate rigidity (h2 close to unity), capital controls (P1), desired long-term exchange variation and long-term real exchange variation set to zero (∆S^���� = 0 and ∆Z^���� = 0), consistent with PPP and the historical peg of the CFAF.
- Monetary policy rule and reaction function (Equations 11–13):
  - Short-term nominal interbank rate RS_t responds to:
    - 4-quarter-ahead inflation forecast deviations from target,
    - output gap,
    - deviation of risk premium from steady-state (influenced by reserve coverage).
  - Parameter h1 captures degree of interest rate control under the peg (h1 = 1: CB loses control; h1 = 0: full sterilization and control).
  - RS_t depends on its lag, neutral nominal interest rate RS_n, inflation expectations ∆4CPI_t+4, and a policy shock ε_RS,t.
  - Risk premium evolution:
    - PPP_t = d1 PPP_t−1 + (1−d1) PPP_ss − d2 CCR_gap_t + ε_PPP,t
    - CCR gap (currency coverage ratio deviation) evolves with k1 CCR_gap_t−1 + k2 (RS_t − RS_n) + k3 |Z_gap_t| + ε_CCR,t
  - The CCR long-term value chosen at 68 percent (chosen close to value at end 2024).

### F. Model parametrization and calibration
- Parameters and steady-state calibrations:
  - Capital buffer steady-state calibrated as historical long-term average.
  - CCR long-term value set at 68 percent.
  - Regulatory requirement K_t^* modeled as a random walk without drift.
- Empirical calibration draws on prior work (Carlos et al., 2022) and related literature; detailed parameter table is provided in Annex I (not reproduced here).

### G. Key findings and policy-relevant implications (from the model text)
- Model capabilities:
  - Decomposes real variables into trend and gap, captures unobserved credit cycle dynamics, and produces medium-term projections and scenarios.
  - Enables assessment of consistency between real and credit cycles and of macrofinancial trade-offs.
- Policy implications:
  - Banks transmit macroprudential policy to the real economy via capital buffers → credit supply → output.
  - Maintaining positive releasable capital buffers (e.g., CCyB) provides headroom and can mitigate procyclical lending cuts during downturns.
  - WAEMU’s CCyB has been set to zero since creation, and macroprudential instruments have been little used; increasing and maintaining releasable buffers could improve resilience.
  - Given the peg and reserve considerations, BCEAO’s monetary policy reacts to reserve/CCR developments; sterilization and capital controls affect effective policy autonomy.
- Institutional notes:
  - Prudential framework (established in 2018) allows BCEAO to implement macroprudential measures in line with financial stability mandate (Article 9 of BCEAO Statutes).
  - Since 2010 CSF-UMOA has rarely issued recommendations on macroprudential instrument implementation; committees have not always met prescribed frequencies.

*Italic: Source — IMF Selected Issues Paper chapter "1. Model Based Simulations and the Actual Data" (sipea2025073).*

### 16.      A mix of Bayesian estimation and calibration is used in setting parameter values

### 16.      A mix of Bayesian estimation and calibration is used in setting parameter values

### Parameterization and estimation approach
- The business cycle part of the model relies on a calibration exercise similar to Carlos et al. (2022).
- Newly introduced parameters for the financial block were first calibrated using regression analysis and then estimated based on Bayesian estimation techniques.
- Model structural parameters were estimated with Bayesian estimation techniques using observable variables over 2004Q1 to 2024Q3.
- Bayesian estimation steps:
  - Construct likelihood by employing Kalman filtering.
  - Combine priors (based on previous studies, e.g., Carlos et al., 2022) with data information; estimate the mode of posterior distributions by maximizing the log posterior function.
  - Use the Metropolis-Hastings algorithm to obtain full posterior distributions and evaluate the marginal likelihood.

### Calibration of steady-state and shock variances
- Steady-state parameters are calibrated to reflect either target levels of variables or historical data averages.
- Calibrated parameters in non-structural equations (primarily trends) are set to achieve gradual adjustment or smooth dynamics toward steady states.
- Standard deviations of shocks are generally calibrated unless explicitly indicated as estimated.
- Calibration is done to achieve reasonable dynamics of gaps and trends when filtering historical data.

### Model simulation and pseudo-out-of-sample forecasting
- Out-of-sample forecasting exercise begins in 2010Q1.
  - Assumption: all foreign and unobserved variables (based on full-sample information) are observed throughout the forecast period.
  - All other variables are known until the quarter preceding each forecast realization over the next 8 quarters.
- Figure 1 (model simulations vs actual) indicates the model has good forecasting capabilities; projected values generally do not diverge from observations when shocks are not large.
- Differences between forecasts and actuals can reflect policy mistakes (temporary deviations in monetary or macroprudential policies from model-prescribed stabilization rules).

### Historical decompositions: business cycle and credit cycle
- Business cycle (output gap) decomposition (Figure 2):
  - Output gap became negative during 2011 (Ivorian crisis and Sahel drought), moved around zero, rose until the COVID-2019 pandemic, and turned negative again during the pandemic.
  - Post-pandemic recovery driven by favorable monetary conditions, gradual recovery of foreign output, and credit demand.
  - Other unmodeled factors (e.g., large fiscal package in 2020) show up as large positive shocks in mid-2020.
  - Feedback from financial sector shocks to the business cycle is limited historically; may reflect absence of financial crisis or significant financial stress in WAEMU during the sample period.
- Credit gap decomposition (Figure 3):
  - Positive credit gap narrowed since late 2023 and is projected to become somewhat negative by end-2024.
  - Positive contributions from favorable business cycle are outweighed by:
    - Strong inertia of the credit gap.
    - Capital buffers being below steady-state level.
    - Credit shocks capturing demand shocks and other unmodeled credit determinants.
  - Capital buffers above steady-state and persistence of credit were major contributors to positive credit gap from 2010 to 2018.

### Impulse response findings
- Business cycle appears to lead the financial cycle; historically limited feedback from financial shocks to the broader economy.
- Impulse responses to a 1 percent shock to output, credit, and capital (Figure 4) — key results:
  - Positive demand shock:
    - Raises output gap and rate of inflation.
    - Boosts real credit growth by 1.6%.
    - Contributes to capital accumulation via increased profitability.
    - Calls for an increase in real interest rate (nominal rate hike greater than rise in inflation); tightening dampens output and reduces credit demand; medium-term return to potential.
    - Credit responses incorporate macro feedback effects through increased capital buffers.
  - Positive credit shock:
    - Allows banks to accumulate capital via higher earnings from above-trend credit growth.
    - Tightening monetary conditions and regulatory discretionary actions put downward pressure on credit supply; credit growth turns negative after one quarter.
    - Credit shocks do not directly affect inflation in the model.
    - Higher capital requirement lowers additional capital buffer difference between bank-held capital and regulatory threshold, raises marginal lending costs, and reduces credit volume.
    - Impact on output is mild; output gap turns slightly negative, reflecting weak credit–real economy link due to limited financial deepening.
    - Results imply a relatively easy trade-off for macroprudential regulator when responding to excessive credit growth and systemic risk.

### Model-based projections and conditioning assumptions
- Figure 5 shows model predictions for inflation, GDP growth, CCR, nominal credit growth, and capital adequacy ratio, conditioned on forecasts for foreign inflation, output growth, and interest rates (IMF WEO October 2024 projections referenced for conditioning).
- Model differs from WAEMU team baseline in Staff Report because:
  - Model focuses only on specified relations; lacks a full external sector and does not encompass expected developments in hydrocarbon production, current account improvements, and reserve accumulation.
  - A CCR target close to end-2024 values was chosen to avoid a monetary policy reaction to CCR; projections are reasonable if expected large-scale hydrocarbon production is not considered.
- Projection highlights:
  - Macroeconomic conditions would call for normalization of monetary policy as FX reserve position improved in late 2024, inflation expectations decline, and credit gap is negative.
  - Negative credit gap expected to close gradually through forecasting period.
  - Credit demand expected to recover in line with activity with a lag; credit-to-GDP ratio to grow again as financial deepening resumes.
  - Bank profitability to improve and gradually restore capital buffers, but capital buffers remain below historical long-term average.
  - Inflation (quarterly) projected to converge to the target level by in early 2025.

### Policy implications and recommendations
- Model links macrofinancial shocks, cyclical risk, and capital buffers — useful for capital requirement policy discussions.
- Current historically low capital buffers relative to steady-state may be insufficient to withstand adverse domestic and international shocks.
- Recent deterioration in financial soundness indicators and pockets of vulnerabilities:
  - Overall capital adequacy ratio declined from 13.7 percent at end-2023 to 13 percent in June 2024 (regulatory norm 11.5 percent).
  - NPLs and restructured loans increased from 8.7 percent at end-2023 to 9 percent in June 2024.
  - Liquidity conditions deteriorated.
- Recommendation: further strengthen capital buffers to mitigate risk amplifiers (concentration, contagion) and emerging risks (interest rate risk), given sovereign exposures and reliance on BCEAO refinancing.
- WAEMU Council of Ministers decision (December 2023) to double minimum paid-in capital for banks to CFAF 20 billion will help system resilience.
- Model also links capital buffers and financial deepening:
  - Banks near regulatory minimum face choice: restrict credit or increase capital buffers; restriction could hinder growth and investment.
  - Current credit penetration is relatively low, at about one-fourth of GDP.
  - Baseline model projects gradual recovery of capital buffers toward steady-state, suggesting financial deepening could be limited under current capital projections.
  - Heavy reliance on central bank refinancing poses risks; greater BCEAO financing to achieve deepening could increase financial stability risks for BCEAO.
- Macroprudential recommendations:
  - BCEAO should be ready to adjust macroprudential policy in line with financial cycle developments.
  - Consider moving toward a positive neutral level for the CCyB and develop a CCyB decision framework in parallel.
  - Gradual phase-in of higher capital requirements during financial expansion can mitigate excessive credit growth and strengthen banks’ loss-absorption capacity.
  - Use the model for simulation exercises to assess macrofinancial impacts of micro- and macroprudential capital requirement changes.

### Conclusions on model usefulness and extensibility
- Model formalizes linkages between real and financial sectors, provides realistic forecasts, and helps assess risks and support monetary and financial policy analysis.
- Model can be used to:
  - Analyze historical macroeconomic and financial sector data.
  - Interpret current disequilibria.
  - Conduct scenario analyses and examine empirical relevance of frictions.
  - Make forecasts to inform policy decisions.
- Financial block is purposely simple; overall model is flexible for future extensions:
  - With minor modifications and additional data, model can address sector-specific financial risks.
  - Can be tailored to resemble an FSAP stress test using supervisory data.
  - Can evaluate macroprudential measures such as the CCyB.

### Annex I — Selected estimated and calibrated parameter values (as presented)
- Estimated Parameters (Prior distribution; Posterior Parameters Value):
  - 훼1: 훽(0.5,0.04) → 0.45
  - 훼2: N(0.45,0.04) → 0.43
  - 훼3: 훽(0.25,0.02) → 0.24
  - 훽1: 훽(0.72,0.06) → 0.69
  - 훽2: N(0.15,0.01) → 0.14
  - 훽3: N(0.28,0.03) → 0.26
  - 훽4: 훽(0.4,0.04) → 0.39
  - 훽5: N(0.1,0.01) → 0.1
  - 훾1: 훽(0.4,0.04) → 0.51
  - ℎ1: 훽(0.4,0.04) → 0.32
  - i1: 훽(0.65,0.06) → 0.69
  - i2: N(0.4,0.03) → 0.41
  - i3: N(0.3,0.03) → 0.3
  - i4: N(0.02,0.001) → 0.02
  - k1: 훽(0.94,0.1) → 0.98
  - k2: N(0.6,0.04) → 0.62
  - k3: N(0.35,0.04) → 0.34
  - 훽(0.9,0.09) → 0.94 (label in source: 휌휌_BBBBBB symbol)
- Calibrated Parameters (as listed in Annex):
  - h2 → 0.97
  - P1 → 0.5
  - g2 → 1
  - g3 → 1
  - g4 → 1
  - ∆S_t^����� → 0
  - ∆Z_t^����� → 0
  - ∆4CCPM_t^��������� → 2
  - ∆4CCPM_t^*����������� → 2
  - PPPPP_sess → 2
  - CC_ssg → 7
  - BBBBBB_sess → 2
  - RSSS_sess1 → 5
  - CCR_sess → 68
  - GGGPP_gPGggh_sess → 5.7

Note: The Annex presents parameter names and values using the original notation from the source; entries are reproduced here exactly as presented.

*Source: IMF staff estimates, sipea2025073 (chapter 16).*

---


_Source: https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025073.pdf_